After entrepreneurs Claudio Reiter and Leon Busch built their first two fully automated biotech labs, the Helsinki-based pair decided to go one better: they instructed an AI agent to complete another on its own.
“We thought it was a bit arrogant to assume humans know best how to build the lab that will teach AI wet-lab science [where scientists work with physical chemicals]. Why not let AI build its own? So we captured what we learned and now an agent can manage a lab build itself,” says Reiter.
The agent is co-ordinating a budget of roughly $500,000 (€455,000), handling everything from procuring instruments to instructing contractors. Other agents are running simple undergraduate-level experiments for their human masters, who plan to ultimately base the company between San Francisco and Zurich.
Reiter believes that within months “the complexity of the experiments they can do will rise until they can meaningfully contribute to scientific research and make discoveries that humans never could”.
READ MORE
Such arrangements are becoming more common and offer a glimpse of how AI is upending start-up culture, traditionally characterised by large teams working long hours and financed by venture capital. Today, a company can be started without either capital or employees. The innovation cycle in start-ups, always rapid, now looks faster than ever – especially for companies whose own products are AI-related. Founders can use AI agents to write code and handle almost any role, from registering a company to designing the product and making marketing videos.
“It’s never been easier to start a company,” says Ziv Reichert, a partner at London-based venture capital firm Phoenix Court.
OpenAI chief executive Sam Altman remarked in 2024 that in a “group chat with my tech CEO friends there’s this betting pool for the first year that there is a one-person billion-dollar company.”
Founders say the game-changing moment for start-ups happened late last year, when new tools such as Anthropic’s Claude Opus 4.5 and its Claude Code made coding cheaper, faster and easier than ever before.
Toby Mather, co-founder of AI start-up Rig, which helps groups integrate company data into reports, apps and AI tasks, wrote in a February newsletter titled “Opus Babies”: “Opus 4.5 appeared on November 24 [2025]. We have since completed 18 months’ of ‘old world’ development in three months.”
[ The end of maths as we know it? Seismic impact of AI has mathematicians worriedOpens in new window ]
Using the new tools, even people who cannot write code can develop software and build products. More and more, entrepreneurs simply speak aloud to AI agents, telling them what to create.
Helena Gagern, who cofounded a tech hub and runs a VC fund for European start-ups called The Embassy in San Francisco, was stunned when she recently caught up with an entrepreneur whose new start-up already had a drug in animal testing – but not a single employee.
“He left his company with the insight, ‘I think I can rebuild every single thing that humans are doing in my company right now with AI’,” she says. “Everything – in terms of disease to work on, the research, the trial design – was driven by AI agents. That was for me the eye-opening moment in terms of how far you can push building a company with practically zero humans working on it.”
She adds that while many start-ups in San Francisco are still raising large sums from venture capitalists, “we’ve seen companies make tens of millions within the first year of operating without raising VC money”.
Vincent Huguet, who in 2013 cofounded Malt, a European platform on which freelance AI engineers and data scientists advertise for work, says that just as computer hardware stopped being a barrier to entry when the cloud arrived 20 years ago, now software is not a barrier either. “It can be made by anyone who is a bit geeky,” he notes.
Isabel Greenslade, the New York-based co-founder of Structured AI, which uses AI to review architectural and engineering drawings, recalls suggesting a possible product to potential customers.
“We recorded the conversation, we put it in Claude Code and just said, ‘Build their idea.’ It can literally mock up a prototype. Two days later you can come back to them with exactly [the product] they discussed.”
Making prototypes or even products almost instantaneously takes some of the risk out of a start-up, says Jimmy Cockerton, founder of Impact Advantage, a platform that measures companies’ social value. “You can use AI to create a good enough version to say, ‘This is what it would look like,’ so people can see and feel and touch it, and it can feel like it’s more advanced and built than it really is, which enables you to test and get feedback.”
There is also early evidence AI is making start-ups more geographically diverse. A recent study by academics at the University of British Columbia and the Stockholm School of Economics, which examined how AI is changing entrepreneurship in the US, found the average “innovative” company was more likely than before to have been created away from tech hubs such as San Francisco or New York.
A Brookings Institution report last year found that while the Bay Area remains dominant for AI, other centres such as Houston and San Antonio in Texas, and Miami and Gainesville in Florida qualify as “star hubs” with high growth in AI jobs. Brookings also singled out Detroit, Nashville, Tampa and Pittsburgh as fast risers.
Use of AI is also fuelling a rise in start-ups. In the 15 years to 2019, the average number of weekly applications to form new businesses in the US never reached 80,000, but since the launch of ChatGPT in November 2022 it has almost always been above 100,000, according to data analysed by asset manager Apollo.
[ Innovators to watch: The long road to building an Irish AI start-upOpens in new window ]
In the UK, the number of new companies incorporated in the application software sector jumped 40 per cent to 28,100 in the first half of the year, making the sector the most popular for start-ups, according to an analysis of Companies House statistics by data group Beauhurst and NatWest.
In London, one repeat founder has launched a start-up staffed only by himself and his partner. Each man has five screens above his desk on which AI agents constantly complete complex tasks.
Saving on wages was only part of the lure, he says. “We asked ourselves, ‘Do we need other humans? Won’t they just slow us down?’” He is also glad not to have to engage in frequent meetings.
However, he warns the pace at which his agents produce, including overnight, can be exhausting for people who have to keep checking the agents’ work and giving them new tasks. Reichert, at Phoenix Court, agrees: “I can synthesise way more information than I could before, so there’s way more decisions that I can be taking. I’ve never been busier.”
Even though the use of agents keeps staffing numbers in check at some start-ups, many will eventually require employees, argues Malt’s Huguet, because products, websites and marketing materials made by AI can be mediocre and samey and can result in so-called “slop companies”.
“The overall level of the playing field has risen, but once you’ve reached the basic level, the difference will be made by the people in your team,” he says. New companies will also need persuasive salespeople to woo clients who are being deluged by spam from AI agents.
AI has added an extra competitive edge to the already cut-throat world of start-ups, given the ease with which rival companies can be created.
Joan Burkovic, chief executive of GetMint, a French company that tracks companies’ visibility on AI platforms, says a business now needs to renew itself constantly: “Our product and technology evolve several times a week,” he says. “A company’s rise can be very fast. But the fall can be all the faster.” – Copyright The Financial Times Limited 2026






















